Determinants of Gold Reserves: An Empirical Analysis for G-7 Countries

Abstract

In this article, we explore the determinants of gold reserves by central banks of Group of Seven (G-7) countries. They are also the suppliers of key currencies. We analyse a sample spanning 24 years period from 1990 to 2014, using panel regression model. Firstly, we compare the patterns of gold reserves in the central banks of G7 with the determinants of their central banks’ total reserves, and also with non-gold international reserves. We find that the factors that affect gold reserves are utterly different from other two models. Secondly, we show that an increase in G-7 countries’ GDP, and also in their exports of goods and services, have positive and significant effects on gold reserves, while Population, Net FDI Liabilities, and Current Account Balance have negative effects on it. Finally, we predict that high economic growth and rising exports and services of G-7 are more likely to lead to an increase in their gold reserves which are effecting 64% of net global wealth.

Full text

Procedia Economics and Finance 38 ( 2016 ) 8 – 16
2212-5671 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Organizing Committee of ICEF 2015.
doi: 10.1016/S2212-5671(16)30172-1
ScienceDirect
Available online at www.sciencedirect.com
Istanbul Conference of Economics and Finance, ICEF 2015, 22-23 October 2015, Istanbul,
Turkey
Determinants of Gold Reserves: An Empirical Analysis for
G-7 Countries
Beyza OKTAYa
*
, Hakan ÖZTUNÇb, Z. Vildan SERİNc
aBeyza OKTAY, Fatih University, 34500, Istanbul, Turkey
bHakan Öztunç, Fatih University, 34500, Istanbul, Turkey
cZ. Vildan SERİN, Fatih University, 34500, Istanbul, Turkey
Abstract
In this article, we explore the determinants of gold reserves by central banks of Group of Seven (G-7) countries. They are also the
suppliers of key currencies. We analyse a sample spanning 24 years period from 1990 to 2014, using panel regression model.
Firstly, we compare the patterns of gold reserves in the central banks of G7 with the determinants of their central banks’ total
reserves, and also with non-gold international reserves. We find that the factors that affect gold reserves are utterly different from
other two models. Secondly, we show that an increase in G-7 countries’ GDP, and also in their exports of goods and services,
have positive and significant effects on gold reserves, while Population, Net FDI Liabilities, and Current Account
Balance have negative effects on it. Finally, we predict that high economic growth and rising exports and services of G-7 are
more likely to lead to an increase in their gold reserves which are effecting 64% of net global wealth.
© 2015 The Authors. Published by Elsevier Ltd.
Peer-review under responsibility of the International Strategic Management Conference.
Keywords: Gold Reserve; G-7; Panel Regression; Central Banks; international reserves.
© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Organizing Committee of ICEF 2015.

9
Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
1. Introduction
Historically gold is accepted as safe haven by investors, households and governments during the economic crises.
According to the World Gold Council, there are multiple different reasons why central banks invest in gold: (1)
Gold acts as a hedge against inflation. (2) Central banks get income from leasing gold. (3) Gold offers worldwide
confidence. (4) Gold offers a diversification benefit. (5) Changes in the world monetary system do not affect gold.
(6) Gold provides physical safety in case other assets are blocked on accounts. (7) The price of gold is unaffected by
bad government decisions, unlike currencies that are prone to bad decisions made by governments (Bernard
Dierinck, 2012) (https://www.gold.org/sites/default/files/documents/gold-investment-
research/liquidity_in_the_global_gold_market.pdf)
In the aftermath of global crisis, the focus shifts into financial assets those can be considered as safe heavens. In
2015, the United States’ central bank kept almost 72 percent of its total monetary holdings as gold reserves. (
http://www.statista.com/statistics/216086/gold-reserve-percentage-in-central-banks-worldwide/ ). Gold could help
investors to suffer losses from inflation (Bolgorian & Gharli, 2010). Gold reserve positions come into light in those
traumatic times. According to Parisi and Diaz (2007), gold had proved to be the most effective commodity for cash
return during the stock exchange crisis in year 1987 and Asian Crisis in 1997. From the perspective of investors and
central banks gold positions and gold reserves are still significant and debatable issues. Yet, it is commonly seen that
after the critical phases of crisis, gold prices are affected negatively - it falls. This issue brings up some questions:
“Do central banks affect gold prices?”, “Are they hidden actors?”
Central banks do some interventions in the financial markets since 1990s by gold sales and gold lending to
manipulate gold prices (Speck 2013). Beside the fact that gold reserve levels increase and existence of various
studies about the factors affecting gold reserve, there is still a gap of agreement about those mentioned factors. Still,
it is a common acceptance that, at the times of global turbulence, gold is seen as a safe investment, a potential
hedge. (Aizenman and Inoue 2013) Many from Marx and Nietzsche to Jack Kemp and Ronald Reagan have seen
gold as a secure anchor and insisted on its role as a foundation of a stable economy. Furthermore Marx had a fear if
gold disappears and money loses weight, there would be destabilizing effect on economy (Taylor 2008).
There are many studies analysing the factors affecting the demand for the international reserves for many different
sample of countries and for many different time periods. Literature review section aims to create an empirical
perspective to the subject. With the light of all these past literature, it is aimed to analyse the factors affecting gold
reserve levels of central banks of the G-7 (Group 7 countries).
2. Literature Review
The increase in the gold reserve positions of key economies is an ongoing and crucial debate. Some economies have
a tendency to make gold stocks either openly or in strict confidence. The numbers given in the literature about the
gold reserve holdings of central banks are corroborative in this content. Kazakhstan’s gold moves, Germany’s recall
for its gold reserves abroad, Mexico’s and Thailand’s action to buy gold are all worth to take attention. Gold is
considered as a very strong competitor to the euro and dollar for being the second biggest reserve instrument.
While the gold is gaining that much importance, there is little consensus in the literature on what factors affects the
countries’ gold reserves. Most of the studies are related with the international reserve levels in general. According to
Cheung, financial crisis is another significant determinant that affects of international reserves (Cheung and Ito
2009)

10 Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
Another research by Aizenman and Lee (2005) is concluded with the support of the idea that precautionary demand
is more important in international reserve hoarding behaviour than the mercantilist view.
Also another comparison is made between the small island economies and emerging market economies, and
questioned whether reserves are too low in the small island economies or is it too high in the developing economies
by the study of Mwase (2012). It is briefly mentioned that the economies with relatively high import shares tend to
hold more reserves since the vulnerability to current account shocks can be considered as an important determinant.
Aizenman and Sun (2009) explains the depletion of international reserves for the period between 2008 and 2009.
Olokoyo, Osabuohien and Salami (2009) analysed foreign reserve and some macro econometric variables in Nigeria
in the period of 1970 and 2007 with annual data. Further empirical studies are done about 10 Asian economies for
the period of 1980 to 2004 with annual data in order to analyze the demand for international reserves. The
explanatory variables were per capita GDP in log, average propensity to import, Exchange rate volatility, volatility
of international reserve holding and financial openness. Panel based regression for these variables is run by Cheung
and Qian (2007). The results were telling that beside the psychological reasons, it is good to hold more reserves in
order to decrease vulnerabilities to speculative attacks and also to enhance growth.
The most inspirational study for our analysis is about central banks tendency to report international reserves
valuation excluding gold positions. Gold holding intensity is seen correlated with “global power”. Additionally, it is
concluded that central banks under-report their gold positions matched with their loss aversion and their wish to
maintain sizeable gold position when gold prices decrease (Aizenman and Inoue 2013)
3. Panel Data Analysis
In panel data the same cross-sectional unit is surveyed over time. Panel data enrich empirical analysis in ways that
might not be possible if it is used cross-section or time-series data. The two types of panel data are generally used
for econometric analyses: fixed effects and random effects panel data models. Both of them can be represented as
below:
ܻ
௜௧ ൌߚ
ଵ௜௧൅ߚଶ௜௧ܺଶ௜௧ ൅ߚ
ଷ௜௧ܺଷ௜௧ ൅ڮ൅ߚ
௞௜௧ܺ௞௜௧ ൅݁
௜௧ǡ݅ ൌ ͳǡʹǡǥǡܰǢݐ ൌ ͳǡʹǡǥǡܶሺͳሻ
Where N is the number of individuals, T is the number of time period. Each element has two subscripts in equation
(1); i stand for group identifier and t refers to tth time period. This property provides to obtain different types of
models in panel data. Fixed effects (FE) assumes that individual group / time have different intercept in the
regression equation (1), while random effects (RE) have the hypothesis that individual group /time have difference
disturbance. In FE, it is suggested that the intercepts of the individuals might be different and these differences could
be due to special features of each individual. FE assumes the coefficients of the repressors’ do not vary across over
time. If both FE and RE turn out significant this is whether to use as a model prototype, then Hausman test is needed
to make a decision for the model. Therefore, there would be two scenarios in the Hausman test: (1) if both FE and
RE models generate consistent point estimates of the slope parameters, they will not differ meaningfully. (2) If the

11
Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
orthogonality assumption is violated, the inconsistent RE estimates will significantly differ from their FE
counterparts (Baum, 2006; Hausman & Taylor, 1981). The Hausman test uses as a -statistic with degrees of freedom
k where k is dimensionality of the estimators.
3.1. Method
A panel regression is based on combination of time series and cross-section data that in this study reveals the link
between the gold reserves and economic indicators. A balanced panel of 175 observations from 7 countries over the
1990–2014 time-span periods (25 years) is used in this study. The sample of the countries represents major
industrial economies: Canada, France, Germany, Italy, Japan, United Kingdom, and USA. There are several
estimation techniques in panel data regression, however the two most remarkable are fixed and random effects in a
panel regression; we use and determine only fixed effects. The data were obtained annually from databank of the
World Bank.
The reserves will be used as dependent factors while macro-economic variables, trade related variables and financial
related variables will be used as independent factors of the reserves. The effects of these independent variables on
the reserves are studied using the following equations:
Model 1: ܴܶ௜௧ ൌߚ
ଵ௜௧൅ߚଶ௜௧ܺଶ௜௧ ൅ߚ
ଷ௜௧ܺଷ௜௧ ൅ߚ
ସ௜௧ ܺସ௜௧ ൅݁
௜௧ǡ݅ ൌ ͳǡʹǡǥǡܰǢݐ ൌ ͳǡʹǡǥǡܶ
Model 2: ܶܩ௜௧ ൌߙ
ଵ௜௧൅ߙଶ௜௧ ܺଶ௜௧ ൅ߙ
ଷ௜௧ܺଷ௜௧ ൅ߙ
ସ௜௧ܺସ௜௧ ൅ݑ
௜௧ǡ݅ ൌ ͳǡʹǡǥǡܰǢݐ ൌ ͳǡʹǡǥǡܶ
Model 3: ܩ௜௧ ൌߛ
ଵ௜௧൅ߛଶ௜௧ ܺଶ௜௧ ൅ߛ
ଷ௜௧ܺଷ௜௧ ൅ߛ
ସ௜௧ܺସ௜௧ ൅݉
௜௧ǡ݅ ൌ ͳǡʹǡǥǡܰǢݐ ൌ ͳǡʹǡǥǡܶ
Where ߚଵ௜௧ǡߙ
ଵ௜௧ǡߛ
ଵ௜௧ are constant terms; ߚଶ௜௧ ǡߚ
ଷ௜௧ǡߚ
ସ௜௧ǡߙ
ଶ௜௧ǡߙ
ଷ௜௧ǡߙ
ସ௜௧ǡߛ
ଶ௜௧ǡߛ
ଷ௜௧ǡߛ
ସ௜௧ are coefficient vectors;
݁௜௧ǡݑ
௜௧ǡ݉௜௧ are the error terms. ܴܶ௜௧ǡܶܩ௜௧ and ܩ௜௧ are dependent variables for each model is as Total Reserves, Total
reserves without gold and gold reserves, respectively. The macro economic variables are selected as GDP (as current
USD) and population which are represented as ܺଶ௜௧ in the equations. The trade related variables are imports of goods
and services, and exports goods and services which are denoted as ܺଷ௜௧ in the equations. The financial related
variables are selected as net foreign direct investment (FDI) liabilities, financial openness index, current account
balance and private capital flows which are embedded as ܺସ௜௧ in the equations.

12 Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
Table 1. Descriptive Statistics
Variable
Mean
Standard
Deviation
Minimum
Maximum
Observations
Total Reserves
overall
1.81e+11
2.60e+11
1.38e+10
1.30e+12
N=175
between
2.04e+11
3.67e+10
6.15e+11
n=7
within
1.78e+11
-3.55e+11
8.61e+11
T=25
Total Reserves excluding gold
overall
1.29e+11
2.54e+11
1.14e+10
1,26e+12
N=175
between
2.08e+11
3.57e+10
6.00e+11
n=7
within
1.64e+11
-3.99e+11
7.87e+11
T=25
Gold Reserves
overall
5.23e+10
7.34e+10
4.55e+07
4.35e+11
N=175
between
5.76e+10
9.64e+08
1.69e+11
n=7
within
5.02e+10
-4.51e+10
3.18e+11
T=25
GDP
overall
3.59e+12
3.64e+12
5.75e+11
1.74e+13
N=175
between
3.59e+12
1.04e+12
1.13e+13
n=7
within
1.46e+12
-1.75e+12
9.69e+12
T=25
Population
overall
1.01e+08
8.11e+07
2.78e+07
3.19e+08
N=175
between
8.69e+07
3.15e+07
2.86e+08
n=7
within
8066144
6.41e+07
1.33e+08
T=25
Imports of Goods and Services
overall
23.41337
8.438562
6.866705
40.0424
N=175
between
8.0337
11.49083
32.56097
n=7
within
3.946026
13.6698
33.4223
T=25
Exports of Goods and Services
overall
23.73086
9.629593
9.002096
45.92429
N=175
between
9.159993
10.7479
34.17721
n=7
within
4.516278
11.06405
36.63393
T=25
Private Capital Flows
overall
0.0616909
3.732165
-10.25906
19.60354
N=175
between
1.328874
-1.616656
2.369808
n=7
within
3.522318
-10.50715
19.35544
T=25
Net FDI Liabilities
overall
1.69423
1.96373
-0.3971911
10.78906
N=175
between
1.148281
0.13121
3.516119
n=7
within
1.649107
-1.152186
10.93855
T=25
Financial Openness Index
overall
2.310687
0.368873
0.1292457
2.389668
N=175
between
0.1186962
2.139158
2.389668
n=7

13
Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
within
0.3520253
0.3007744
2.561197
T=25
Current Account Balance
overall
-0.1822378
2.794546
-5.912297
7.535958
N=175
between
2.154257
-3.00677
2.583894
n=7
within
1.951609
-4.369098
4.888888
T=25
3.2. Finding and discussions
The descriptive statistics and regression and post estimation results are provided by the construction of the FE
model. Stata 13.1 software program was used to compile data for the panel regression analysis.
The descriptive statistics are shown in Table 1. The analysis started with finding means and standard deviations for
cross-sectional time series data. The variations could be seen between and within countries.
According to test results in Table 2, our three models can be constructed significantly. On the other hand, some
differences seem in the significance of the independent variables in the models. In model 1, the six variables; GDP
current USD, Population, Imports of Goods and Services, Net FDI Liabilities, Financial Openness Index, and
Private Capital Flows are found to be significant for all countries in the study. The variables explain approximately
65% of variations in total reserves of our selected 7 countries. The results for Exports of Goods and Services and
Current Account Balance are found to be insignificant in explaining the first dependent variable the total reserves of
these countries. Population, Imports of Goods and Services and GDP current USD are found to be significant with
the values (p=0.099, p=0.036, and p=0.093, respectively) and have positive effects on these countries’ total reserves.
Net FDI Liabilities, Private Capital Flows and Financial Openness Index are also found to be significant with the
values (p=0.000), (p=0.014), (p=0.050), respectively. However, their signs of coefficient estimates reveal that they
affect the total reserves negatively.
In model 2, the five variables; Population, Imports of Goods and Services, Net FDI Liabilities, Financial Openness
Index, and Private Capital Flows are found to be significant for all countries in the study. Approximately 43%of
variations in Total reserves without gold can be explained by the independent variables. On the other side, GDP
current USD, Exports of Goods and Services, and Current Account Balance do not make any effects on the Total
reserves without gold. Population, Imports of Goods and Services have positive effects on these countries’ total
reserves without gold while Net FDI Liabilities, Financial Openness Index and Private Capital Flows have negative
effects.
In model 3, the five variables; GDP current USD, Population, Exports of Goods and Services, Net FDI Liabilities,
Current Account Balance are found to be significant for the countries in the study. These variables explain

14 Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
approximately 46% of variations in gold reserves of our selected 7 countries. Significance and insignificance of
variables in model 3 are utterly different from other two models. Imports of Goods and Services, Financial
Openness Index and Private Capital Flows reveal insignificant results while they are all significant in other two
models. GDP current USD and Exports of Goods and Services are found to be significant with the values (p=0.001)
and (p=0.020), respectively. However the signs of estimates show that GDP current USD and Exports of Goods and
Services affect Gold reserves growth positively while Population, Net FDI Liabilities, and Current Account Balance
have negative effects on it. Population, Net FDI Liabilities, and Current Account Balance are also found to be
significant with the values (p=0.001), (p=0.003), and (p=0.001), respectively. Population has an inverse relationship
to gold reserves, and its effectiveness rate is very high.
Table 2. Model Estimation Results for Countries
Model 1: Total Reserves
Model 2: Total Reserves
without Gold
Model 3: Gold Reserves
Independent Variables
Coefficients
Prob.
Values
Coefficients
Prob.
Values
Coefficients
Prob.
Values
GDP current USD
0.3766616
0.093
0.1970849
0.304
1.178352
0.001
Population
2.811502
0.099
1.454323
0.001
-12.92231
0.001
Imports of Goods and Services
2.279917
0.036
1.053521
0.026
-0.221241
0.841
Exports of Goods and Services
-1.058281
0.399
1.207873
0.127
3.631261
0.020
Net FDI Liabilities
-0.0589504
0.000
-0.0391202
0.008
-0.1140747
0.003
Financial Openness Index
-0.1318324
0.050
-0.1678508
0.043
0.0545913
0.588
Current Account Balance
0.0329385
0.384
0.0575601
0.151
-0.2421766
0.001
Private Capital Flows
-0.0837414
0.014
-0.0847545
0.036
0.0404209
0.364
Constant
-39.69334
0.133
-70.97534
0.003
-0.1140747
0.000
Number of Observations
175
175
175
Adjusted R-square
0.6493
0.4338
0.4626
From Table 3, F tests (F (8, 24) =132.68, p=0.000<0.01, F (8, 24) =164.87, p=0.000<0.01, F (8, 24) =49.51,
p=0.000<0.01) are for fixed effects and LM tests (X2=0.000, p=1.000>0.1, same for three models) are for random
effects. As a double check to decide for fixed effects model, Hausman tests are performed for each models and the
null hypothesis of the Hausman tests are not rejected. Then the preference would be for the fixed effects model for
all three models.

15
Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
Table 3. The Fixed Model Tests for Countries
Model 1: Total Reserves
Model 2: Total Reserves
without Gold
Model 3: Gold Reserves
Test Name
Test Statistics
Prob. Values
Test Statistics
Prob. Values
Test Statistics
Prob. Values
LM ( Var (u)=0,Lambda=0)
0.00
1.000
0.00
1.000
0.00
1.000
F(8,24)
132.68
0.000
164.87
0.000
49.51
0.000
Hausman
1144.03
0.000
308.95
0.000
100.95
0.020
Serial Correlation :
ALM(Lambda=0)
535.9
0.000
226.16
0.000
278.17
0.000
Modified Wald test for group wise
heteroscedasticity
320.93
0.000
240.22
0.0000
311.11
0.0000
Modified Bhargava et al. Durbin-
Watson
0.36084485
---
0.25490407
---
0.4562052
---
Baltagi-Wu LBI
0.51791015
---
0.40772534
---
0.63386741
---
Breusch-Pagan LM test of
independence
111.023
0.0000
112.194
0.0000
76.88
0.0000
For testing the heteroscedasticity and auto correlation assumptions in Fixed Effects, Modified Wald test and
Modified Bhargava et al. Durbin Watson (Baltagi-Wu LBI) tests are used. Modified Wald tests show
heteroscedasticity. Both Modified Bhargava et al. Durbin Watson and Baltagi-Wu LBI test values are below 2 in all
models. This means that autocorrelations exist in our three models. Also Breusch-Pagan LM test results show that
the correlation between countries exists. To overcome these problems, Driscoll-Kraay fixed effects model is
appropriate that Table 2 is shaped according to this model results.
Table 4 gives the rankings of countries, from high to low, based on the models’ fixed effect values. The countries’
total reserve fixed effect values in ranking according to Driscoll-Kraay fixed effects model. Based on the first
model’s results; Canada has the most fixed effect value of the total reserves where USA has the least. Model 2 gives
very similar results for total reserves without gold with Model 1. Even there would be slightly differences in
ranking, the top and the bottom of the list is the same as the first model. The last model reveals very dissimilar
results with the previous two tables. It gives nearly reverse results with the others. This time USA is at the top of the
list but Canada is at the end.
Table 4. Ranking of the Countries According to Models
Countries
Model 1
Fixed Effects
Ranking
Model 2
Fixed Effects
Ranking
Model 3
Fixed Effects
Ranking
Canada
1.4
1
4.17
1
18.63
7
Italy
0.73
2
1.13
2
7.51
5
Japan
0.6
3
1.127
5
0.77
2
France
0.37
4
0.63
4
-2.24
4
United
Kingdom
0.06
5
-0.21
3
-3.24
6
Germany
-0.28
6
-0.312
6
-5.07
3
USA
-2.95
7
-6.54
7
-16.36
1

16 Beyza Oktay et al. / Procedia Economics and Finance 38 ( 2016 ) 8 – 16
4. Conclusion
G-7 has been focused initially on macroeconomic policy coordination and transnational issues since 1975. The
central banks of G7 turned net buyers of gold since 2010. Therefore, we explored the determinants of gold holdings
of G-7 for a sample spanning a 24 years period from 1990 to 2014. Firstly we found that the factors that impact on
gold reserves were different from that other total international reserve and also non-gold international reserve
holdings. Not just gold but also other international reserve levels tend to move along the results of the crises. In
order to explain for total reserve these six variables are tested: GDP current USD, population, imports of goods and
services, net FDI liabilities, financial openness index, and private capital flows. All of the G-7 countries are found
significant in this research. The variables explain approximately 65% of variations in total reserves of our selected
seven major industrial countries. Precautionary motive forces central banks to accumulate reserves but mercantilist
approach is another reason which holds sizable portion of the literature. Reserves are mostly seen as a buffer stock
or a hedge to the crisis’ effects. Besides, gold has a power to stabilize economies. Secondly our results indicate that
both increase GDP current USD and exports of goods and services affect an increase in gold reserves while
population, net FDI liabilities, and current account balance have negative effects on it. These findings are consistent
with the previous literature. There is strong relationship between gold reserves and state power. That’s why gold is
accepted as a safe haven and strategic asset by the countries during the financial crises. Finally the countries are
ranked using the results of the fixed effects model. The findings show USA has the most fixed effect value of the
total reserves where Canada has the least. We predict that high economic growth and rising exports and services of
G-7 are more likely to lead to an increase in their gold reserves which are affecting 64% of net global wealth.
References
Aizenman, J. and J. Lee (2005). "International reserves: precautionary vs. mercantilist views, theory and evidence."
Aizenman, J. and K. Inoue (2013). "Central banks and gold puzzles." Journal of the Japanese and International Economies 28: 69-90.
Aizenman, J. and Y. Sun (2012). "The financial crisis and sizable international reserves depletion: From ‘fear of floating’to the ‘fear of losing
international reserves’?" International Review of Economics & Finance 24: 250-269.
Baltagi, B. H. 2005. Econometric analysis of panel data (3rd ed.). West Sussex, UK: John Wiley & Sons.
Baum, C. F. 2006. An introduction to modern econometrics using Stata, Stata press.
Baur, D. and T. McDermott (2009). Is Gold a Safe Haven? International Evidence. IIIS Discussion Paper, Institute for International Integration
Studies: 38.
Bolgorian, M. and Gharli, Z. (2011). A multifractal detrended fluctuation analysis of gold price fluctuations. Acta Physica Polonica B. 42(1),
pp.159-168. doi:10.5506/APhysPolB.42.159.
Cheung, Y.-W. and H. Ito (2009). "A cross-country empirical analysis of international reserves." International Economic Journal 23(4): 447-481.
Cheung, Y. W. and X. Qian (2009). "Hoarding of international reserves: Mrs Machlup's wardrobe and the Joneses." Review of International
Economics 17(4): 824-843.
Dierinck, Bernard (2012). Determinants of the Gold Price,p.3 retrieved from: http://lib.ugent.be/fulltxt/RUG01/001/893/290/RUG01-
001893290_2012_0001_AC.pdfHausman, J. A. &
Taylor, W. E. 1981. Panel data and unobservable individual effects. Econometrica: Journal of the Econometric Society, 1377-1398
Mwase, N. (2012). "How Much Should I Hold? Reserve Adequacy in Emerging Markets and Small Islands."
Olokoyo, F. O., et al. (2009). "Econometric Analysis of Foreign Reserves and Some Macroeconomic Variables in Nigeria (1970–2007)*."
African Development Review 21(3): 454-475.
Parisi, A., Parisi, F. and Diaz, D. (2008). Forecasting gold price changes: Rolling andrecursive neural network models. Journal of Multinational
FinancialManagement. 18, pp.477–487.
Speck, D. (2013). The Strange Behaviour of Gold during Crises. The Gold Cartel, Palgrave Macmillan UK: 8-15.
Taylor, M. C. (2008). Confidence games: Money and markets in a world without redemption, University of Chicago Press.